TL;DR

Dario Amodei has rejected the argument that his warnings about AI risk are what turned the public against the technology. Responding to investor Gavin Baker, who urged him to advocate more positively for the industry, Amodei accepted that public opinion is a serious problem but located the cause elsewhere: a long-running collapse of trust in companies, governments and the technology sector.

The exchange

Baker’s case, made on a podcast and on X, was that dire warnings from AI executives have fed hostility in the United States, particularly toward data centres, and that Amodei has lost the regulation argument. Amodei answered that his output has been roughly balanced between risk and benefit, and that he wrote Machines of Loving Grace precisely because the industry was not describing an inspiring enough future.

What he would not accept is the causal claim. Ordinary people, in his account, assume institutions are devising some fresh way to take advantage of them, and the current backlash is only the latest expression of a suspicion decades in the making.

The concession worth noting

He offers a sharper self-criticism than his critic did. The fair charge against AI companies, his own included, is that the big promises remain undelivered — and that this, rather than messaging or marketing, is where the argument should sit. Promising a cure for cancer, he says, has become a cliché; curing cancer would change minds.

On regulation he rejects the framing that rules necessarily mean regulatory capture and concentrated power, noting that outside Silicon Valley many people read regulation as a constraint on corporate power. Anthropic’s proposals, he says, are built to slow frontier labs while advantaging smaller competitors. Open weights help with concentration but do not solve it, on his reading, because they shift power toward whoever holds the most compute and chips.

The UK relevance

This lands as UK organisations face their own version of the trust question with staff and customers rather than regulators. The argument Amodei is making — that credibility follows delivered outcomes, not better communication — applies directly to internal AI rollouts, where scepticism is usually a rational response to tools that were announced with more confidence than they earned.

Looking forward

A frontier lab chief conceding that his industry’s central problem is undelivered value is a useful marker. It is also a testable one, and the deadline for that test is entirely of the industry’s own making.